Counterfactual Prediction for Online Content Attribution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing online systems struggle to accurately predict the likelihood of users causing desired events when presented with content, as they do not account for external influences or content from sources outside the system.
Innovation Solution
The online system receives and processes three data sets – explicit user data, explicit event data, and implicit user data, implicit event data – to generate counterfactual and causative rates, using machine learning models to estimate the likelihood of users causing events when not presented with content within the system, and combines these rates to provide a more accurate prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the online system uses traditional prediction methods to estimate user actions, then the prediction process is simple, but the prediction accuracy is low because external influences are not accounted for
Solution Approach 1:
The prediction process is segmented into multiple independent components: counterfactual rate prediction (users not shown content), causative rate prediction (users shown content), and external influence estimation. Each component processes specific data features separately before combining results, allowing complex external factors to be broken down into manageable segments that can be processed independently.
Solution Approach 2:
The patent introduces an intermediary counterfactual rate prediction component that estimates what would happen without online system content. This intermediary prediction acts as a mediator to isolate and measure external influences, allowing the system to subtract the counterfactual rate from the total observed rate to get the true causative effect of online system content.
2Measurement precision
If the online system collects and processes multiple data sets including implicit and explicit data, then the prediction accuracy improves, but the data processing complexity increases
Solution Approach 1:
Different data features are processed with different levels of detail and complexity appropriate to their specific characteristics. Explicit user data receives different processing than implicit user data, and different feature sets are created for counterfactual versus causative predictions. This local quality approach ensures each data type gets the appropriate level of processing without uniformly increasing complexity across all data.
Solution Approach 2:
The patent transforms multiple data sets into a multi-dimensional feature space where explicit and implicit data are combined into composite features. This dimensional transformation allows the system to process complex multi-source data by projecting it into a unified feature dimension that can be handled by the prediction models, effectively managing complexity through dimensional organization.
3Measurement precision
If the online system uses counterfactual prediction to account for external influences, then the accuracy of attributing conversions correctly improves, but the computational requirements increase
Solution Approach 1:
The counterfactual rate prediction is performed as a preliminary step before final conversion attribution. By pre-computing the counterfactual rate that represents external influences, the system avoids repeated heavy computations during the actual attribution process. This preliminary action caches the external influence estimate that can be reused across multiple attribution decisions.
Solution Approach 2:
The system changes parameters by using estimated aggregate features rather than processing individual user data points for every prediction. The counterfactual rate and causative rate are computed using aggregated feature sets that summarize user behavior patterns, reducing the computational parameters from individual user-level detail to population-level statistics that require less processing energy.
Data Source
AI summary
An online system receives explicit user data and explicit event data, and implicit user data and implicit event data from a third party system. The online system generates an implicit users/implicit events data feature, an explicit users/explicit events data feature, and an explicit users/implicit events data feature. The online system generates a prediction of the counterfactual rate based on the implicit users/implicit events data feature, the explicit users/explicit events data feature, and the explicit users/explicit events data feature, the counterfactual rate indicating the likelihood that target users matching certain characteristics caused an event to occur when the target are not been presented with content by the online system, the content configured to induce users to cause the event to occur. A combined prediction rate is presented to the third party system based on the counterfactual rate.